For the social impacts analysis, social life cycle assessment (UNEP
2020) was used, a relatively recent methodology (the first methodological version was published in 2009) for which, unlike environmental LCA (E-LCA), there is still no reference standard (ISO/AWI 14075 is under development) but only generic guidelines (Iofrida et al.
2018). However, since the basic assumption of S-LCA is to adapt LCA and its standards to the social dimension, the three mandatory steps required for E-LCA, namely
goal and scope definition,
life cycle inventory (LCI), and
life cycle impact assessment (LCIA), were followed in this study and will be explained in detail in the following paragraphs. The literature proposes different methodologies for conducting social impact assessment. In particular, Richter et al. (
2023), identified that the best-known approaches are the Social Footprint Method (SFM) (McElroy
2008), SeeBalance (Schmidt et al.
2004), and S-LCA (UNEP
2020), which covers about 80% of social sustainability studies internationally. Then there are a further plethora of approaches, including, just a few, contingent valuation method (CVM) (Nautiyal and Goel
2021), social return on investment (SROI) (De Leon
2021), most significant change (MSC) (Henry et al.
2022), labor process theory (LPT) (Braverman
1998) or multi-criteria decision analysis (MCDA). However, some reasons mainly related to practicality and applicability to the FD context, relevance to the study, and resources have led to the preference for S-LCA over these methodologies. For example, the SFM, which describes a context-based approach to measuring, managing, and reporting an organization’s social sustainability performance, covers only a few issues, risking underestimating the problem. SeeBalance, on the other hand, covers just socio-efficiency, while not considering pure social sustainability. The CVD, being mostly based on interviews to determine willingness to pay to estimate the economic value of goods with no market value, could be susceptible to doubt and lead to wide variability in the quality of results, as also noted by Whitehead and Haab (
2013). The S-ROI, an impact evaluation framework that measures the intangible social values used in both for-profit and nonprofit types of institutions, is sometimes found to be too subjective and judgmental (Gibbon and Dey
2011), difficult to quantify (Lowe
2013), as well as susceptible to risk in overlooking some context and process elements. MSC is a method used to integrate evaluations of the outcomes and impact of a given program through mini-narratives that contextualize the effect a program has on individuals. However, it was excluded because it was inconsistent with the objective of the study. LPT, on the other hand, is a Marxist approach in the study of production relations in industrial capitalism (Gandini
2019), in which managers seek to control the way work is organized, the pace and duration of work (because it is decisive for profitability), and could focus exclusively on the worker-employer relationship, while not considering the social implications along the entire value chain, thus excluding a broader group of stakeholders as in the case of S-LCA. In fact, by using S-LCA, one could investigate not only the conditions of the workforce but also broader social ramifications thus ensuring a more inclusive evaluation than the relatively narrow focus of work process theory. Finally, MCDA methods are a collection of systematic approaches developed specifically to support the evaluation of alternatives in terms of multiple and often conflicting objectives (Krainyk et al.
2021), seeking to identify the choice whose consequences would imply greater social development value. Decision analysis theory is designed to help an individual or group in choosing among a set of pre-specified alternatives. While particularly useful when comparing different social, economic, and environmental indicators (Zanghelini et al.
2018), MCDA evaluation methods could consider many more subjective elements when evaluating different criteria (e.g., stakeholder opinions), as also confirmed by Myllyviita et al. (
2017) and Torre et al. (
2024).
In addition, the choice of S-LCA was also supported by additional, no less important reasons: First of all, since the objective of this research is to provide a broad assessment of the social dimensions of food delivery, S-LCA was deemed the most suitable methodology due to its systematic framework, which allows for a more structured analysis, encompassing various social aspects, reaching as far as consumer protection. Then the use of life-cycle approaches (LCA, LCC, and S-LCA) as a tool to promote the shift to sustainable patterns of production and consumption is increasingly recognized internationally. In particular, S-LCA currently has broad scientific consensus and is the most widely considered approach in the literature for assessing social sustainability, as also extensively documented by Tragnone et al. (
2022) and Richter et al. (
2023). This, then, could lead, for example, to easier comparability between similar studies. But also, because, in addition to being particularly useful in the broader range of CSR tools, the S-LCA framework is distinguished by its multiple connections with global initiatives (Agenda 2030, 10-Year Framework of Programs on Sustainable Consumption and Production, the International Labor Organization’s Decent Work Agenda, Guiding Principles for Business and Human Rights) (UNEP
2020), thus aiding in their pursuit. Moreover, the above is also associated with the fact that recently, several research needs have been expressed, including the application of S-LCA in case studies (Ramos Huarachi et al.
2020), which further motivated this study. Finally, a concluding element that was particularly important in choosing S-LCA among the many social impact assessment methodologies was the possibility of being able to be guided by the PSILCA database (Eisfeldt and Ciroth
2020), which provided internationally recognized threshold values and guidelines that could then return results that were reproducible and as comparable as possible with other future studies, in other contexts, and for other products or services. Thus, while recognizing the presence of various methodologies for assessing social impacts, the decision to use S-LCA was influenced primarily by the context and objective of the study (to provide an overview of the social sustainability of FD in Italy), its proven applicability in similar contexts, its ability to effectively integrate stakeholder perspectives according to a holistic view, as well as its robustness and acceptability in the international field, also under the forthcoming ISO 14067.
3.1 Goal and scope definition
The main purpose of S-LCA is to provide decision support, which can create an effect if decision-makers follow the outcome of the assessment and choose the alternative with the most favorable social consequences (Di Cesare et al.
2018). In the UNEP guidelines, social impacts are interpreted either as consequences due to specific behavior engaged in by one or more stakeholders, as a downstream effect of socioeconomic decisions, or concerning attributes possessed by an individual, group, or society. Therefore, consistent with UNEP, the S-LCA goal of this study is to consider the potential negative social impacts within the FD sector in Italy, to highlight the specific behavior of one or more stakeholders, and then present the results according to different levels of risk for different groups of stakeholders. In this way, the study could then help to measure the possibility of negative effects occurring in the FD sector and then avoid them through one or more preventive actions. As far as the functional unit (FU) is concerned, since social impacts are intangible, in S-LCA studies, they are not dependent on physical flows (Wei et al.
2022) and are not necessarily proportional to them, but reflect the influence of the sector’s behavior toward various stakeholders (Zamagni et al.
2021). Therefore, although ideally, we could refer to a unit of food packaged and delivered in Italy, for a conceptual issue, i.e., both because the social impacts might not be related on a linear scale and because of the variability of the foods in FD, the FU is not defined. This choice is also consistent with other literature on S-LCA studies. These include, for example, Mulyasari et al. (
2023), who in a study related to the social impacts of palm oil show that because it is difficult to relate intangible social impacts to physical flows, they do not consider a FU. Even Umair et al. (
2015), who, in their study regarding the management of all ICT e-waste entering the informal recycling sector in Pakistan, using only qualitative data on social impacts, do not express impacts by FU, which only serves to specify the scope of the assessment. And finally, Macombe et al. (
2018), call the introduction of functional units in S-LCA studies a “
stubborn insistence.” This is because the very relationship between the quantity of FUs and the magnitude of impacts is purely linear, that is, the magnitude of impacts caused by n FUs is n times the magnitude of impacts caused by one FU. In contrast, on the other hand, for example, union freedom does not depend on the number of shoes sold. Regarding system boundaries, the S-LCA, being still evolving, shows some methodological weaknesses and shortcomings, including difficulties in defining system boundaries in detail, as also pointed out by recent literature studies (Tragnone et al.
2022). Indeed, since this is a sectoral study and not a product study, it might be difficult to enclose FD in a well-delineated boundary. Verisimilarly, the geographical boundaries of the sector include the Italian state since the study is about FD in Italy. From a process perspective, ideally, the study could be
gate-to-gate, as it considers the social impacts of workers during the transport phase (so when the product leaves the restaurant or point of sale, excluding the food preparation phase) until the moment of consumption, excluding end-of-life.
3.3 Social life cycle impact assessment
Depending on the different objectives of those carrying out the study, from a methodological point of view, in S-LCA studies, there are two families of social life cycle impact assessment approaches: One assessment with respect to a reference scale, called reference scale assessment (RS S-LCIA), also known as Type I and another assessment that considers a cause-and-effect relationship called impact pathway assessment (IP S-LCIA), also known as Type II (Sureau et al.
2020). The latter estimates the implications of different pathways to an endpoint, thus enabling the assessment of long-term impacts through various correlations (Orola et al.
2022). However, IP S-LCIA is more difficult to apply, mainly because of the obstacles associated with defining cause-and-effect mechanisms in the social context, turning out to be in fact scarcely used, as also confirmed by literature studies (Tragnone et al.
2022; Zafar et al.
2024). The Type I approach, on the other hand, relies on data, information, or judgments and provides an immediate assessment of performance and risks (e.g., at the inventory indicator level) and not the further propagation of effects, effectively not establishing a link between activity and long-term impacts. Instead, based on available information, it estimates the likely magnitude and importance of potential social impacts. For this reason, the collected data are compared with performance reference points (PRPs) (e.g., the number of hours per week worked by the individual worker is compared with the 40 h per week defined by the convention of the International Labor Organization). Therefore, RS S-LCIA provides for the development of a reference scale with explicitly defined levels for each inventory indicator, collected in the LCI phase (UNEP
2020). The reference scales can be represented in three main ways: (1) in non-numeric terms (colors, letters, tick marks); (2) with linear scores, where each level of the scale corresponds to a value above the previous one and thus to better performance; and (3) with non-linear scores, where each level of the scale is assigned, a customized value based on the gap to be defined between two levels of the scale. Within this study, the method chosen to convert the inventory collected in the previous stage into potential social impacts is the Type I approach. The reasons for this choice are as follows: First, it is considered the most current and widely used approach in case study applications (Mármol et al.
2023), as well as being a particularly straightforward measure. This is consistent with the aim of the research, which is not to delve into the long-term effects of FD in Italy, but to focus, on the basis of readily available data, on the social risks related to the behavior of the actors and organizations involved in FD, providing an overview of the state of the art in the relevant context. Secondly, S-LCIA Type I has a close link to social reporting approaches, such as corporate social responsibility standards (Sureau et al.
2020). This could be particularly important in the food delivery sector, where companies are expected to report on their social responsibilities, especially when a poorly protected category, such as riders, is involved. This could be further true, especially considering aspects such as the fair treatment of riders, their rights and wages, as well as safety standards and other social implications related to their operations. This connection could then serve as a basis for assessing and improving the social sustainability of FD. In addition, RS-LCIA is in line with major S-LCA databases (Bouillass et al.
2021), such as PSILCA, which is an important support for this study.
Finally, since the Type I approach allows for the evaluation of all stakeholder groups and their subcategories, it is more consistent with the multi-actor perspective discussed in this paper. At this point, for the evaluation of the collected data, PRPs were defined for all quantitative indicators to estimate social risk compared to some international standards, local legislations, labor laws, etc. Specifically, the PRPs considered in this research are the same as those proposed by PSILCA, and they vary according to the various indicators, with maximum and minimum values, with ranges divided into five equal parts leading to linear scores and a scale composed of six social risks (Eisfeldt and Ciroth
2020): no risk, very low risk, low risk, medium risk, high risk, and very high risk. These different levels of risk were then represented through a speedometer chart in which each risk corresponds to a different color (respectively: dark green, light green, orange/yellow, red, dark red). As an example, in the case of the subcategory “children in employment,” which defines the category “child labor,” the indicator is the percentage of children aged 7–14 years involved in employment and the PRPs are as follows: 0 (no risk), 0 < y < 2. 5% (very low risk), 2.5% ≤ y < 5% (low risk), 5% ≤ y < 10% (medium risk), 10% ≤ y < 20% (high risk), and 20% ≤ y (very high risk), as well as no data (Eisfeldt and Ciroth
2020). So, for example, if a value should be 3%, then it falls within the PRP 2.5–5%, indicates a low-risk (light green), and so on for each indicator. After that, consistent with UNEP guidelines that propose a point scale (one to five), each risk was given a numerical score from 0 to 5. Specifically: 0 = no risk; very low risk = 1; low risk = 2, medium risk = 3; high risk = 4, very high risk = 5, and the results were finally expressed through a radar chart, also in agreement with other studies, such as Tokede et al. (
2020) and Gompf et al. (
2022). A more complete overview of reference scales and PRPs can be found in the supplementary materials (Table
S2). Regarding stakeholder categories, i.e., groups of people involved in the value chain of the product system (Bouillass et al.
2021), four were considered, consistent with UNEP guidelines. Specifically: workers, value chain actors, society, and local community. Eleven risk categories were then chosen: child labor, forced labor, fair salary, working time, discrimination, health and safety, social benefits, legal issues, workers’ rights, promoting social responsibility, health and safety, and migration. Finally, the subcategories chosen are 13 (Children in employment, frequency of forced labor, sector average wage per month, hours of work per employee per week, gender wage gap, fatal accident at workplace, provision of protection for employees, evidence of violations of laws and employment regulations, trade union density, membership in an initiative that promotes social responsibility along the supply chain, violations of mandatory health and safety standards, presence of management measures to protect consumer health and safety, migrant workers in food delivery), each of which in turn is expressed by indicators. The reason behind the choice of these categories is twofold and can be attributed to both the availability of the data and their fit with the food delivery industry and the various impact categories. The stakeholders, risk categories, and subcategories, as well as the various indicators, are described in the following paragraphs.
3.3.2 Value chain actors
Then the data were normalized by considering whether in the ESG the company treats the topic thoroughly (there is no risk of unsustainable business practices) or superficially or poorly (there is the risk of unsustainable business practices). S-LCA does not provide in an industry-wide manner the ability to provide quantitative estimates to assess the extent to which social responsibility is taken seriously and ensured by companies within specific industries, but it might be useful to implement a qualitative approach.